This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
diagnostics
Review
A Systematic Review of Asthma Phenotypes Derived byData-Driven Methods
Francisco Cunha 1, Rita Amaral 2,3,4,5,* , Tiago Jacinto 2,4, Bernardo Sousa-Pinto 2,3,6 and João A. Fonseca 2,3,7
1 Faculty of Medicine, University of Porto, 4200-319 Porto, Portugal; [email protected] Center for Health Technology and Services Research (CINTESIS), Faculty of Medicine, University of Porto,
3 Department of Community Medicine, Information and Health Decision Sciences (MEDCIDS),Faculty of Medicine, University of Porto, 4200-319 Porto, Portugal
4 Department of Cardiovascular and Respiratory Sciences, Porto Health School, Polytechnic Institute of Porto,4200-072 Porto, Portugal
5 Department of Women’s and Children’s Health, Paediatric Research, Uppsala University,751-05 Uppsala, Sweden
6 Basic and Clinical Immunology Unit, Department of Pathology, Faculty of Medicine, University of Porto,4200-319 Porto, Portugal
7 Allergy Unit, CUF Porto Hospital and Institute, 4100-180 Porto, Portugal* Correspondence: [email protected]; Tel.: +351-9-1700-6669
Abstract: Classification of asthma phenotypes has a potentially relevant impact on the clinicalmanagement of the disease. Methods for statistical classification without a priori assumptions(data-driven approaches) may contribute to developing a better comprehension of trait heterogeneityin disease phenotyping. This study aimed to summarize and characterize asthma phenotypesderived by data-driven methods. We performed a systematic review using three scientific databases,following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria.We included studies reporting adult asthma phenotypes derived by data-driven methods usingeasily accessible variables in clinical practice. Two independent reviewers assessed studies. Themethodological quality of included primary studies was assessed using the ROBINS-I tool. Weretrieved 7446 results and included 68 studies of which 65% (n = 44) used data from specializedcenters and 53% (n = 36) evaluated the consistency of phenotypes. The most frequent data-drivenmethod was hierarchical cluster analysis (n = 19). Three major asthma-related domains of easilymeasurable clinical variables used for phenotyping were identified: personal (n = 49), functional(n = 48) and clinical (n = 47). The identified asthma phenotypes varied according to the sample’scharacteristics, variables included in the model, and data availability. Overall, the most frequentphenotypes were related to atopy, gender, and severe disease. This review shows a large variabilityof asthma phenotypes derived from data-driven methods. Further research should include morepopulation-based samples and assess longitudinal consistency of data-driven phenotypes.
Asthma is one of the most common chronic diseases in the world and its prevalenceis increasing due to the continuous expansion of western lifestyle and urbanization [1].Asthma is a chronic inflammatory disease of the airways, characterized by at least partiallyreversible airway obstruction and bronchial hyper-responsiveness [1,2]. Global Initiativefor Asthma (GINA) currently defines asthma as a heterogeneous disease, with a historyof respiratory symptoms that vary over time and in intensity, together with variableexpiratory airflow [2]. Taking into account that asthma is such a heterogeneous condition
with complex pathophysiology, phenotypic classification is essential for the investigationof etiology and treatment tailoring [3].
Patients with asthma have been categorized into subgroups using theory- or data-driven approaches. In the classical theory-driven approach, patients with asthma areclassified in categories defined a priori according to current knowledge (e.g., based onetiology, severity, and/or triggers) [4]. However, this approach generates asthma pheno-types that are not mutually exclusive, and the correlation with therapeutic response andprognosis might not be the most adequate [5].
On the other hand, the data-driven (or unsupervised) approach, which is unbiasedby previous classification systems, often starts with a broad hypothesis and uses relevantdata to generate a more specific and automatic hypothesis, providing an opportunityto better comprehend the complexity of chronic diseases [4]. Several classes of data-driven algorithms have been involved in tackling the issue of trait heterogeneity in diseasephenotyping. The techniques most used to address phenotypic heterogeneity in healthcare data include distance-based (item-centered, e.g., clustering analysis) and model-based(patient-centered, e.g., latent class analysis) approaches, both of which are not mutuallyexclusive [6].
Distance-based approaches use the information on the distance between observationsin a data set to generate natural groupings of cases [3]. The most commonly used clusteringanalysis methods are hierarchical, partitioning (k-means or k-medoids), and two-stepclustering, which can be roughly described as a combination of the first two. Hierarchicalclustering analysis functions by creating a hierarchy of groups that can be represented in adendrogram, while the partitional methods divide the data into non-overlapping subsetsthat allow for the classification of each subject to exactly one group [3].
On the other hand, the most used model-based approaches, which use parametricprobability distributions to define clusters instead of the distance/similarities between theobservations [7], are latent class analysis (LCA), latent profile, and latent transition analysis.
Despite the existence of studies that identified clusters mainly coincident with otherlarger-scale cluster analyses [8–10], there is a lack of consistency of phenotypes and appliedmethods. Therefore, this systematic review aimed to summarize and characterize asthmaphenotypes derived with data-driven methods in adults, using variables easily measurablein a clinical setting.
2. Materials and Methods
In this systematic review, we followed the Preferred Reporting Items for System-atic Reviews and Meta-Analyses (PRISMA) statement [11] and the Patient, Intervention,Comparison and Outcome (PICO) strategy [12] to improve the reporting of this system-atic review.
2.1. Search Strategy
Primary studies were identified through electronic database search in PubMed, Scopus,and Web of Science (first search in August 2020; updated in March 2021). Broad medicalsubject headings (MeSH) and subheadings, or the equivalent, were used and search queriesare presented in Table 1.
Diagnostics 2021, 11, 644 3 of 63
Table 1. List of queries used for searching online databases.
Database Research Query
Pubmed
(phenotyp*[Title/Abstract] OR cluster*[Title/Abstract])AND (“Asthma”[MeSH] OR asthm*[Title/Abstract])
AND (“Adult”[MeSH] OR “Adult” [Title/Abstract] ORadult*[ Title/Abstract] OR “Middle
Aged”[Mesh:NoExp] OR “Aged”[Mesh:NoExp]) AND(humans[mesh:noexp] NOT animals[mesh:noexp]) NOT
((Review[ptyp] OR Meta-Analysis[ptyp] ORLetter[ptyp] OR Case Reports[ptyp]))
Scopus
(TITLE-ABS-KEY (asthm*) AND TITLE-ABS-KEY((phenotyp* OR cluster*)) AND TITLE-ABS-KEY((adult* OR “middle aged” OR elderly))) AND(EXCLUDE (DOCTYPE, “re”) OR EXCLUDE
(DOCTYPE, “le”) OR EXCLUDE (DOCTYPE, “ed”) OREXCLUDE (DOCTYPE, “no”) OR EXCLUDE
(DOCTYPE, “ch”) OR EXCLUDE (DOCTYPE, “sh”))
Web of Science
(TS = (asthm*) AND TS = ((phenotyp* OR cluster*))AND TS = ((adult* OR middle aged or elderly))) NOTDT = (BOOK CHAPTER OR REVIEW OR EDITORIAL
MATERIAL OR NOTE OR LETTER)
2.2. Study Selection
Studies were considered eligible when reporting asthma phenotypes determined bydata-driven methods in adult patients (≥18 years old), exclusively using variables easilyavailable in a clinical setting. We did not apply exclusion criteria based on language orpublication date criteria. Studies using genotyping variables were excluded.
Two authors (F.C. and R.A.) independently screened all the identified studies by titleand abstract, after excluding duplicates. Subsequently, potentially eligible studies wereretrieved in full-text and assessed independently by two authors, who selected those thatmet the predefined inclusion and exclusion criteria. Disagreements in the selection processwere solved by consensus. Non-English publications were translated if considered eligible.
Cohen’s kappa coefficient was calculated to evaluate the agreement between the tworeviewers in the selection process.
2.3. Data Extraction
Two authors (F.C. and R.A.) were involved in data extraction. Study design, setting,inclusion criteria, patients’ characteristics, variables, and data-driven methods used forphenotyping, and the obtained phenotypes, were assessed for each study.
Variables were divided into eight domains for simplicity and practicality of analysis(Table 2).
Diagnostics 2021, 11, 644 4 of 63
Table 2. List of variables covered by each domain.
Domain Variables
PersonalGender, age, smoking, BMI, family history ofasthma, race, education level, socioeconomic
status
FunctionalFEV1, FVC, FEV1/FVC, KCO or other lung
function measurements, reversibility ofobstruction, bronchial hyperresponsiveness
Clinical
Symptoms, exacerbations, asthma control,asthma severity scores, activity limitation, age
of onset, disease duration, work-relatedasthma, near-fatal episode, associated
comorbidities, imaging-related
AtopyAtopic status, serum IgE, sensitization,
allergen exposure, rhinitis or other allergicdiseases, skin prick test, immunotherapy
Inflammatory FeNO, blood eosinophils, and neutrophils,sputum eosinophils, and neutrophils, hsCRP
MedicationRegular medication, daily dose of prednisolone
or equivalent, use of rescue bronchodilator,oral corticosteroid use
Healthcare use Emergency department use, hospitalizations,stays in ICU, unscheduled visits to GP
Behavioral
Attitude towards the disease, perception ofcontrol, observed behavior, psychological
status, confidence in doctor, stress in daily life,impact on activities in daily life
Body mass index (BMI), forced expiratory volume in 1 s (FEV1), forced vital capacity (FVC), carbon monoxidetransfer coefficient (KCO), immunoglobulin E (IgE), fractional exhaled nitric oxide (FeNO), high-sensitivityC-reactive protein (hsCRP), intensive care unit (ICU), general practitioner (GP).
2.4. Quality Assessment
Two independent researchers (F.C. and R.A.) independently performed the assessmentof the quality of the evidence using the ROBINS-I approach [13]. Based on the informationreported in each study, the authors judged each domain as low, moderate, serious, orcritical risk of bias. Any disagreement was solved by consensus. Quality assessment wassummarized in a risk of bias table.
3. Results3.1. Study Selection
A total of 7446 studies were identified in the literature search, of which 2799 wereduplicates. After screening all titles and abstracts, which resulted in the exclusion of 4472records, 175 citations were determined to be potentially eligible for inclusion in our review.Subsequently, full-text assessment resulted in the exclusion of 107 studies in total, including28 studies incorporating variables or phenotypes with limited applicability in a clinicalsetting or using phenotypes obtained in previous studies, and 17 studies without availablefull text. Unavailable references included meeting abstracts, conference papers, posters,and older studies from local publications with no traceable full text. In the end, 68 studiesof data-driven asthma phenotypes studies were included. A flowchart for study selectionis depicted in Figure 1.
Diagnostics 2021, 11, 644 5 of 63Diagnostics 2021, 11, x FOR PEER REVIEW 5 of 58
Figure 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram illustrating the studies’ selection process.
3.2. Study Characteristics All the 68 studies [8–10,15–79] were published between 2008 and 2020 and recruited
patients mostly from specialized centers (n = 44, 65%). We identified seven population-based studies. The median sample size of all studies was 249 individuals (range 40–7930).
The included primary studies used a wide variety of methods for cluster analysis, with the most common method being hierarchical cluster analysis (n = 19), followed by k-means cluster analysis (n = 16) and two-step cluster analysis (n = 14). Latent class analysis was the most used model-based approach (n = 9) (Figure 2).
Figure 1. Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram illustrating thestudies’ selection process.
For the selection process, the Cohen’s kappa coefficient and the percentage of theagreement were calculated were determined to be 0.76 and 98%, respectively. These resultsindicate substantial agreement [14].
3.2. Study Characteristics
All the 68 studies [8–10,15–79] were published between 2008 and 2020 and recruitedpatients mostly from specialized centers (n = 44, 65%). We identified seven population-based studies. The median sample size of all studies was 249 individuals (range 40–7930).
The included primary studies used a wide variety of methods for cluster analysis,with the most common method being hierarchical cluster analysis (n = 19), followed byk-means cluster analysis (n = 16) and two-step cluster analysis (n = 14). Latent class analysiswas the most used model-based approach (n = 9) (Figure 2).
Diagnostics 2021, 11, 644 6 of 63Diagnostics 2021, 11, x FOR PEER REVIEW 6 of 58
Figure 2. Data-driven method chosen for asthma phenotyping ordered by absolute frequency of use.
It was not possible to retrieve the variables used in two studies [15,16]. The remaining 66 studies of our review were applied a wide range of variables in their respective analysis. Personal variables (e.g., age, gender, BMI, or smoking) were included in the analysis of 74% of the previously mentioned 66 studies. Variables belonging to the lung function, clinical, and atopy domains were all used in more than half of these studies. Figure 3 shows the percentage of studies that used each one of the represented domains of variables.
Figure 3. Proportion of each domain of variables in the 66 studies with retrievable chosen variables.
The characteristics of the 68 studies included in our review are summarized in Table 3.
Figure 2. Data-driven method chosen for asthma phenotyping ordered by absolute frequency of use.
It was not possible to retrieve the variables used in two studies [15,16]. The remaining66 studies of our review were applied a wide range of variables in their respective analysis.Personal variables (e.g., age, gender, BMI, or smoking) were included in the analysis of 74%of the previously mentioned 66 studies. Variables belonging to the lung function, clinical,and atopy domains were all used in more than half of these studies. Figure 3 shows thepercentage of studies that used each one of the represented domains of variables.
Diagnostics 2021, 11, x FOR PEER REVIEW 6 of 58
Figure 2. Data-driven method chosen for asthma phenotyping ordered by absolute frequency of use.
It was not possible to retrieve the variables used in two studies [15,16]. The remaining 66 studies of our review were applied a wide range of variables in their respective analysis. Personal variables (e.g., age, gender, BMI, or smoking) were included in the analysis of 74% of the previously mentioned 66 studies. Variables belonging to the lung function, clinical, and atopy domains were all used in more than half of these studies. Figure 3 shows the percentage of studies that used each one of the represented domains of variables.
Figure 3. Proportion of each domain of variables in the 66 studies with retrievable chosen variables.
The characteristics of the 68 studies included in our review are summarized in Table 3.
Figure 3. Proportion of each domain of variables in the 66 studies with retrievable chosen variables.
The characteristics of the 68 studies included in our review are summarized in Table 3.
Diagnostics 2021, 11, 644 7 of 63
Table 3. Characteristics of the included studies.
Study ID (Author,Year) Setting, Design Inclusion Criteria in
the Analysis
Number of PatientsIncluded in the
AnalysisAge Patients’
Characteristics
Variables Used forCluster Analysis
(Number andDomains)
Method Used forCluster Analysis
Agache, 2018 [17]Single center(Romania),
cross-sectional
Diagnosis of seasonalallergic rhinitis and
asthma57 34.12 ± 10.59
Intermittent asthma:35 (8 were
uncontrolled);Persistent asthma: 22
(10 wereuncontrolled)
11 variables:personal, atopy
K-means ClusterAnalysis
Alves, 2008 [18] Single center(Brazil), cohort
Diagnosis of severeasthma,
treatment-compliant88 56 ± 12
Female: 73%;ICS in high dose:
67%;OCS: 30%;
LABA: 88%
12 variables:personal, functional,
clinical, atopyFactor Analysis
Amaral, 2019 [19]Population-based
(NHANES—USA),cross-sectional
Adults (≥18 years)with current asthma 1059 N.A. N.A.
Moderate to severebronchial asthma, onmaintenance therapyin the last four weeks,
age ≥18 years
40 46.37 ± 14.77 Female: 65%
16 variables:personal, functional,
clinical, atopy,inflammatory
Hierarchical ClusterAnalysis
Zaihra, 2016 [79]
Difficult asthmacohort (Montreal
Chest Institute of theMcGill University
Health Centre,Canada)
Subjects aged 18–80years with moderate or
severe asthma (ATScriteria)
125 (48 moderateasthmatics and 77severe asthmatics)
Moderate asthmatics:46.6 ± 11.2;
Severe asthmatics:49.9 ± 12.6
Female: moderateasthmatics—48%,
severeasthmatics—56%
Personal, functional,clinical,
inflammatory
K-means ClusterAnalysis
Not applicable (N.A.), inhaled corticosteroids (ICS), oral corticosteroids (OCS), long-acting β2 agonists (LABA), Global Initiative for Asthma (GINA), bronchial hyperreactivity (BHR), American Thoracic Society(ATS), forced expiratory volume in 1 s (FEV1), Asthma Quality of Life Questionnaire (AQLQ), forced vital capacity (FVC), World Health Organization (WHO), Spanish Guideline on the Management of Asthma(GEMA), chronic obstructive pulmonary disease (COPD).
Diagnostics 2021, 11, 644 22 of 63
3.3. Asthma Phenotypes
The number of phenotypes per study ranged from two to eight with a median of four,obtained in 23 studies (34%). A majority of studies (82%) identified between three and fivephenotypes. The most frequent phenotypes in our analysis were atopic asthma, severeasthma, and female asthma with multiple variants.
We observed that 36 studies (53%) evaluated the consistency of phenotypes based on atleast one of the following criteria: longitudinal stability, cluster repeatability, reproducibility,and/or validity.
A visual representation of the variables used for phenotyping by each study is por-trayed in Table A1 (Appendix A). Studies with an assessment of consistency are highlighted.
Table 4 represents the defining variables of phenotypes obtained by each study. Thefull phenotypes are compiled in Table A2 (Appendix A). The results are stratified by adata-driven method, and the frequency of phenotypes in the sample is presented foreach study.
In hierarchical cluster analysis, the most frequent phenotypes were atopic/allergicasthma, mentioned 24 times in 13 studies, and late-onset asthma, mentioned 19 times in12 studies. A common association with atopic asthma was the early age of onset, whilelate-onset asthma was recurrently linked with severe disease. Atopic asthma was also themost frequent phenotype in two-step cluster analysis. In both k-means and k-medoidscluster analysis, severe asthma occurred the most often.
In model-based methods, latent class analysis studies identified mostly phenotypesrelated to symptoms. Factor analysis used severity of disease to classify asthma, while latenttransition analysis used allergic status and symptoms. One study derived longitudinaltrajectories in terms of pulmonary function using latent mixture modeling.
3.4. Risk of Bias Assessment
We used the ROBINS-I tool to assess the risk of bias. The methodological quality ofthe studies was predominantly moderate (n = 29). Of the 68 included studies, 18 wereconsidered to be at overall low risk of bias, while other 18 studies were considered to be atserious risk of bias. Only three studies were judged to be at critical risk of bias. The resultsare portrayed in Table 5.
Diagnostics 2021, 11, 644 23 of 63
Table 4. Characterization of the phenotypes obtained in each study according to the defining variables (column), with each row within each study corresponding to one phenotype.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Hierarchical Cluster Analysis
Baptist, 2018[22]
Late
Mild
Atopic
Severe
Belhassen, 2016[23]
Less medication
Fixed doseinhalers
Freecombination
Bhargava, 2019[15]
Childhood Mild Preserved Atopic
Male Overweight Adolescent Severe Atopic
Female Obese Late Severe Least atop.
Female Obese Young age Mild Atopic
Delgado-Eckert,2018 [30]
Mild/Mod.
Severe
Fingleton, 2015[31]
Mod./Severe Atopic
COPD
Obese
Mild Atopic
Mild Intermittent
Diagnostics 2021, 11, 644 24 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Fingleton, 2017[32]
COPD Late Severe
COPD Early
Atopic
Adult Nonatopic
Early Mild Intermittent Atopic
Khusial, 2017[39]
Early Atopic
Female Late
Reversible
Smokers
Exacerbators
Konno, 2015 [44]
Early Atopic Mild eos
Smokers Late Fixed limitation Intense Th2
Smokers Late Fixed limitation Low Th2
Nonsmokers Late Low Th2
Female Nonsmokers,high BMI Late Intense Th2
Loureiro,2015 [8]
Early Mild Allergic Eosinophilic
Female Moderate Long evolution Allergic Mixed
Female, young Early Brittle Allergic No evidence
Female Obese Late Severe Highly sympt. Mixed
Late Severe Long evolution Chronicobstruction Eosinophilic
Diagnostics 2021, 11, 644 25 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Moore, 2010 [51]
Female, young Childhood Normal Atopic
Female, slightlyolder Childhood Atopic
Female, older
Childhood Severe Atopic
Female Late Less atopy
Nagasaki, 2014[54]
Late Nonatopic Paucigranulocytic
Early Atopic
Late Eosinophilic
Poor control Low FEV1 Mixedgranulocytic
Qiu, 2018 [60]
Female Early Small degree ofobstruction
Sputumneutrophilia
Female Nonsmokers Severe airflowobstruction
High sputumeosinophilia
FemaleModerate
reduction ofFEV1
Sputumneutrophilia
Male Smokers Severe airflowobstruction
High sputumeosinophilia
Sakagami, 2014[63]
Female Low IgE
Young Early Atopic
Older Late Less atopic
Diagnostics 2021, 11, 644 26 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Schatz, 2014 [64]
Female, white Adult Low IgE
Atopy
Male
Nonwhite
Aspirinsensitivity
Seino, 2018 [65]
Elderly Severe Poor control Adherencebarriers
Elderly Low BMI Severe Poor control No adherencebarriers
Younger High BMI Not severe Controlled No adherencebarriers
Sendín-Hernández,
2018 [67]
Mild Intermittent Low IgE Without familyhistory
Mild IntermediateIgE
With familyhistory
Mod./Severe Needs CS andLABA High IgE With family
history
Sutherland, 2012[70]
Female Nonobese
Male Nonobese
Obese Uncontrolled
Obese Controlled
Diagnostics 2021, 11, 644 27 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Weatherall, 2009[75]
SevereChronic
bronchitis +emphysema
Variableobstruction Atopic
Emphysema
Atopic Eosinophilic
Mild obstruction No otherfeatures
Nonsmokers Chronicbronchitis
Ye, 2017 [77]
Early Atopic
Moderate Atopic
Late Nonatopic
Fixedobstruction
Youroukova,2017 [78]
Late Impaired Nonatopic
Smokers Late High sympt.,exacerbations
Aspirinsensitivity Late Symptomatic Eosinophilic
Early Atopic
K-means Cluster Analysis
Agache, 2010[17]
Severe rhinitis Polysensitization
Male Severe rhinitis Exposure to pets
High IgE,polysensit.
Diagnostics 2021, 11, 644 28 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Amelink, 2013[21]
Severe Persistentlimitation Eosinophilic
Female Obese Symptomatic Low sputum eos High health careuse
Mild/Mod. Controlled Normal
Choi, 2017 [27]
Normal airway,increased lungdeformation
Luminalnarrowing,
reduced lungdeformation
Wall thickening
Luminalnarrowing,
increase in airtrapping,
decreased lungdeformation
Deccache, 2018[29]
Confident
Committed
Questing
Concerned
Diagnostics 2021, 11, 644 29 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Gupta, 2010 [16]
Severe Concordantcontrol score Eosinophilic
Greaterbronchodilator
response
Female High BMI Severe High controlscore Low eos
Severe High controlscore Low eos
Severe Low controlscore Eosinophilic
Lee, 2017 [47]
Near-normal
Asthma
COPD
Asthmatic-overlap
COPD-overlap
Musk, 2011 [53]
Male normal
Female normal
Female Obese
Younger Atopic
Male Atopic High eNO
Male Poor FEV1 Atopic
BHR Atopic
Diagnostics 2021, 11, 644 30 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Oh, 2020 [56]
High UA, T.Chol., AST, ALT,
and hsCRPHigh eos
Intermediate
Low UA, T.Chol. and T. Bili.
Park, 2015 [57]
Long duration Markedobstruction
Female Normal
Male Smokers Reduced
High BMI Borderline
Park, 2013 [58]
Smokers
Severe Obstructive
Early Atopic
Late Mild
Rakowski, 2019[61]
Low eos
Intermediate eos
High eos
Rootmensen,2016 [62]
COPD without emphysema
COPD with emphysema
Allergic
Overlap withCOPD Atopic
Diagnostics 2021, 11, 644 31 of 63
Table 4. Cont.
Study ID(Author, Year)
Defining Variables of Phenotypes
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Demographics Comorbidities Onset Severity Symptoms,Treatment Lung Function Atopy Inflammation Others
Janssens, 2012[37]
Well-controlled
Intermediatecontrol
Poorlycontrolled
Latent Mixture Modeling
Park, 2019 [59]
Male, older Smokers Less atopic
Smokers Higher IgE
Younger More atopic
Female Nonsmokers
Studies are stratified by a data-driven method. Phenotypes are compiled in their full extent in Appendix A. Chronic obstructive pulmonary disease (COPD), body mass index (BMI), eosinophils (eos), forcedexpiratory volume in 1 s (FEV1), forced vital capacity (FVC), immunoglobulin E (IgE), corticosteroids (CS), inhaled corticosteroids (ICS), oral corticosteroids (OCS), long-acting β2 agonists (LABA), AsthmaQuality of Life Questionnaire (AQLQ), exhaled nitric oxide (eNO), uric acid (UA), cholesterol (Chol.), bilirubin (Bili.), high-sensitivity C-reactive protein (hsCRP), bronchial hyperreactivity (BHR).
Diagnostics 2021, 11, 644 43 of 63
Table 5. Risk of bias assessment using ROBINS-I.
Study ID (Author, Year) Confounding Selection ofPatients
The studies included in our review were in accordance with most of the Strengtheningthe Reporting of Observational Studies in Epidemiology (STROBE) checklist items [80].
4. Discussion4.1. Main Findings
This systematic review revealed a high degree of variability regarding the data-drivenmethods and variables applied in the models among the studies that identified data-drivenasthma phenotypes in adults. There was a lack of consistency in the studies concerning thestudy setting, target population, choice of statistical method and variables, and ultimately,the label of the phenotype. Overall, the most frequent phenotypes were related to atopy,gender (female), and severe disease.
Different statistical methodologies were applied among the included studies, withhierarchical and k-means clustering being the most common ones. The earliest study in thisreview (2008) applied a two-step clustering approach to two different sets of patients [33].In the group of patients of the primary care setting, three phenotypes were determined,namely, “early-onset atopic asthma”, “obese, non-eosinophilic asthma”, and “benignasthma.” In the group of patients with refractory asthma managed in secondary care, fourphenotypes were obtained “early onset atopic asthma”, “obese, non-eosinophilic asthma”,“early onset symptomatic asthma with minimal eosinophilic disease”, and “late-onset,eosinophilic asthma with few symptoms” [33]. These phenotypes persisted in later studies,with different variants [8,15,42,55].
Most of the studies recruited patients from specialized centers. However, we identifiedtwo population-based studies with a low risk of bias, both using model-based statisticaltechniques [20,25]. Amaral et al. identified different classes of allergic respiratory diseasesusing latent class analysis in a population of 728 adults. The study obtained seven pheno-types, which were distinguished according to allergic status and degree of probability ofnasal, ocular, and bronchial symptoms [20]. Boudier et al. applied latent transition analysiswith nine variables covering personal and phenotypic characteristics on longitudinal dataof 3320 adult asthmatics, determining seven phenotypes characterized by the level ofasthma symptoms, the allergic status, and pulmonary function. These results revealedstrong longitudinal stability [25].
There were four population-based studies with some identifiable validation process.Amaral et al. derived phenotypes independently for two age groups and found similarproportions in both age groups for the two obtained data-driven subtypes (“highly symp-tomatic with poor lung function”, and “less symptomatic with better lung function”), andfor previously defined hypothesis-driven subtypes. However, the set of variables was sub-optimal to differentiate asthma subgroups [19]. Makikyro et al. applied latent class analysisto identify four asthma subtypes in women and three subtypes in men. Phenotypes wereclassified according to the control and severity of the disease. The subsequent addition of aset of covariates verified the accuracy of results [50].
An improvement of the characterization of asthma heterogeneity is an essential stepin the development of more personalized approaches to asthma management and therapy.There is a need for further research to produce population-based studies with analysisof the longitudinal consistency of data-driven phenotypes. Ilmarinen et al. performedclustering on longitudinal data of Finnish patients with adult-onset asthma. Their approachwith 15 variables resulted in the determination of five phenotypes with longitudinalstability, namely “nonrhinitic asthma”, “smoking asthma”, “female asthma”, “obesity-related asthma”, and “early onset atopic adult asthma” [35]. Furthermore, Khusial et al.identified a set of five phenotypes with longitudinal stability in a primary care cohort ofadult asthmatics: “smokers”, “late-onset female asthma”, “early atopic asthma”, “reversibleasthma” and “exacerbators” [39]. Certain similarities with the results of the study byIlmarinen et al. are identifiable.
Diagnostics 2021, 11, 644 46 of 63
Hsiao et al. found a higher risk of asthma exacerbations in current smoker and ex-smoker clusters in males, as well as in atopy and obesity clusters in females [34]. Park et al.observed an association between smoking males and reduced lung function [57].
The most used dimensions were variables regarding personal, clinical, and functionaldata. However, other dimensions were used in several studies. For example, Lefaudeuxet al. demonstrated that clustering based on clinicophysiologic parameters can producestable and reproducible clusters [48]. Deccache et al. aimed to characterize treatmentadherence with a multidimensional approach encompassing asthma control, attitudetowards the disease, and compliance with treatment [29]. Finally, Labor et al. aimed toassess the association of specific asthma phenotypes with mood disorders—five phenotypeswere identified by cluster analysis of cross-sectional data in a sample of adult patients of atertiary center: “allergic asthma”, “aspirin-exacerbated respiratory disease”, “late-onsetasthma”, “obesity-associated asthma”, and “infection-associated asthma” [46].
An ongoing investigation is being conducted to identify novel targets and biomarkersfor a better understanding of the pathophysiology of asthma. Eventually, the broaderavailability of emerging molecular and genetic tools may complement the traditionalclinical variables in the determination of asthma phenotypes [81].
4.2. Strengths and Limitations
We should note that this study has limitations. In an attempt to assemble a completeoverview of data-driven asthma phenotyping, some of the included studies focused onspecific contexts, which hampered their external validity. Another limitation concernsthe possibility of selection bias, as the definition of asthma varied across the studies(questionnaire-based and/or functional-based). This may possibly have implications onselection bias for participant selection and information bias if there are wrong classificationand assessment of participants. Other important limitations concern the low quality ofmost included studies since, of the 68 included studies, 32 did not attempt to assess theconsistency of results, and only 18 were considered to be at low risk of bias. Moreover,the association between the obtained phenotypes and the clinical outcomes was out of thestudy’s scope and should be further explored.
To our knowledge, this is the first systematic review that summarized data-drivenasthma phenotypes, based on easily accessible variables, in adults. Unsupervised methodshave emerged as a novel tool in adult asthma phenotyping, with the advantage of beingfree from a priori biases; this study provides an overview of the current state in thefield, which may be useful to clinical practitioners and researchers, particularly in theunderstanding of the heterogeneity of asthma. The main strength of this review is theexhaustive compilation of asthma phenotypes with a detailed description of the data-drivenmethods used (Appendix A). Additionally, our study included an extensive literaturesearch by applying no language or date restrictions and performing risk of bias assessmentby ROBINS-I tool. The high number of included publications proves the existence of a needto classify asthma patients using data-driven methods due to the limitations of classicaltheory-driven approaches.
In conclusion, data-driven methods are increasingly used to derive asthma phenotypes;however, the high heterogeneity and multidimensionality found in this study suggestthat both clinic and statistical expertise are required. Further research should focus onpopulation-based samples and evaluation of longitudinal consistency of phenotypes.
Author Contributions: Conceptualization, R.A., T.J. and J.A.F.; methodology, F.C., T.J., B.S.-P. andR.A.; software, F.C. and R.A.; validation, F.C. and R.A.; formal analysis, F.C. and R.A.; investigation,F.C. and R.A.; resources, R.A., J.A.F., B.S.-P.; data curation, F.C.; writing—original draft preparation,F.C. and R.A.; writing—review and editing, F.C. and R.A.; visualization, F.C. and R.A.; supervision,R.A.; project administration, R.A. and J.A.F.; funding acquisition, not applicable. All authors haveread and agreed to the published version of the manuscript.
Funding: This research received no external funding.
Diagnostics 2021, 11, 644 47 of 63
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A
Table A1 displays the variable domains used for phenotyping by each study. Studieswith an assessment of phenotype consistency are highlighted.
Table A1. Representation of variables used by each study, stratified by a data-driven method. Studies with an evaluation ofphenotype consistency are marked. Variables are presented in the form of domains: personal (P), functional (F), clinical (C),atopy (A), inflammatory (I), medication (M), health care use (H), and behavioral (B).
Study ID (Author, Year) DomainsP F C A I M H B
Hierarchical Cluster AnalysisBaptist, 2018 [22] x x x x x
Belhassen, 2016 [23] xBhargava, 2019 [15] Variables were not retrieved.
Delgado-Eckert, 2018 [30] xFingleton, 2015 [31] x x x xFingleton, 2017 [32] x x x xKhusial, 2017 [39] x x x x x xKonno, 2015 [44] x x x xLoureiro, 2015 [8] x x x x x xMoore, 2010 [51] x x x x x x
Nagasaki, 2014 [54] x x x x xQiu, 2018 [60] x x x x
Sakagami, 2014 [63] x x xSchatz, 2014 [64] x x x xSeino, 2018 [65] x x x
Sendín-Hernández, 2018[67] x x x x x x
Sutherland, 2012 [70] x x x xWeatherall, 2009 [75] x x x x
Ye, 2017 [77] x x x x x x xYouroukova, 2017 [78] x x x x x
K-means Cluster AnalysisAgache, 2010 [17] x x
Amelink, 2013 [21] x x xChoi, 2017 [27] x
Deccache, 2018 [29] xGupta, 2010 [16] Variables were not retrieved.
Lee, 2017 [47] x x xMusk, 2011 [53] x x x x
Oh, 2020 [56] x xPark, 2015 [57] x x x xPark, 2013 [58] x x x x
Rakowski, 2019 [61] xRootmensen, 2016 [62] x x x x
Tanaka, 2018 [71] xTay, 2019 [72] x x x xWu, 2014 [10] x x x x x x x
Zaihra, 2016 [79] x x x x
Diagnostics 2021, 11, 644 48 of 63
Table A1. Cont.
Study ID (Author, Year) DomainsP F C A I M H B
Two-step Cluster AnalysisHaldar, 2008 [33] x x x xHsiao, 2019 [34] x x x x
Ilmarinen, 2017 [35] x x x x xJang, 2013 [36] x xKim, 2018 [40] xKim, 2017 [41] x x x xKim, 2013 [42] x x x
Konstantellou, 2015 [45] x x xLabor, 2017 [46] x x x x
Lemiere, 2014 [49] x x x xNewby, 2014 [55] x x x x x x x
Serrano-Pariente, 2015[68] x x x x
Wang, 2017 [74] x x x x xWu, 2018 [76] x x x
K-medoids Cluster AnalysisKisiel, 2020 [43] x x x
Lefaudeux, 2017 [48] x x x xLoza, 2016 [9] x x x
Sekiya, 2016 [66] x x x x xLatent Class Analysis
Amaral, 2019 [19] x x x xAmaral, 2019 [20] x x x x x
Bochenek, 2014 [24] x x x x xChanoine, 2018 [26] x
Couto, 2018 [28] x x x x xJeong, 2017 [38] x x x x
Makikyro, 2017 [50] x x x x xSiroux, 2011 [69] x x x x
van der Molen, 2018 [73] xFactor Analysis
Alves, 2008 [18] x x x xMoore, 2014 [52] x x x x x
Latent Transition Analysis//Expectation-maximizationBoudier, 2013 [25] x x x xJanssens, 2012 [37] x x x x
Latent Mixture ModelingPark, 2019 [59] x x
Table A2 summarizes the phenotypes obtained by each study with the respectivefrequency in the sample. The results are stratified by a data-driven method.
Diagnostics 2021, 11, 644 49 of 63
Table A2. Asthma phenotypes in adult patients were derived by data-driven methods in the includedstudies and stratified by the data-driven method applied. The percentage of subjects that belong toeach phenotype is presented when available.
Study ID (Author, Year) Label
Hierarchical Cluster Analysis
Baptist, 2018 [22]
- “Late-onset asthma” (38%)- “Mildest asthma” (22%)- “Atopic, long duration of asthma” (26%)- “The most severe asthma” (14%)
Belhassen, 2016 [23]
- “Low levels of dispensation of controller medication, fewervisits to the GP” (64%)
- “Received fixed-dose combination inhalers” (32%)- “Received free combination of ICS and LABAs” (4%)
Bhargava, 2019 [15]
- “Milder, childhood-onset, atopic, normal weight, preservedlung function” (40%)
- “Polysensitization and severe rhinitis” (53%)- “Male sex, exposure to pets, and severe rhinitis” (19%)- “High total serum IgE and polysensitization” (28%)
- “Low variability in eos levels with low values” (28%)- “Large variability in eos levels with intermediate values”
(20%)- “Smallest variability in eos levels with the highest values”
(52%)
Rootmensen, 2016 [62]
- “COPD patients without signs of emphysema” (17%)- “Patients with emphysematous type of COPD” (27%)- “Patients with characteristics of allergic asthma” (26%)- “Overlap syndrome of atopic asthma and COPD” (30%)
Tanaka, 2018 [71]
- “Rapid exacerbation, young to middle-aged, hypersensitiveto environmental triggers and furred pets” (42%)
- “Fairly rapid exacerbation, middle-aged and older and lowperception of dyspnea” (40%)
- “Slow exacerbation, high perception of dyspnea, smokers,and chronic daily mild-moderate symptoms” (18%)
Tay, 2019 [72]
- “Chinese females with late-onset asthma and the bestasthma control” (42%)
- “Non-Chinese females with obesity and the worst asthmacontrol” (12%)
- “Multi-ethnic with the greatest proportion of atopicpatients” (46%)
Diagnostics 2021, 11, 644 54 of 63
Table A2. Cont.
Study ID (Author, Year) Label
Wu, 2014 [10]
- “Healthy control subjects” (25%)- “Mild asthma” (36%)- “Mostly severe asthma and frequent symptoms, low AQLQ
scores, a high degree of allergic sensitization” (5%)- “Early onset allergic asthma with low lung function and
eosinophilic inflammation” (21%)- “Later-onset, mostly severe asthma with nasal polyps and
eosinophilia” (8%)- “Early onset severe asthma, the most symptoms, the lowest
lung function, frequent and high-intensity health care use,and sinusitis” (6%)
Zaihra, 2016 [79]
- “Severe asthmatics and predominantly late-onset disease”(12%)
- “Female, severe asthmatics, with higher BMI” (14%)- “Severe asthma with reductions in pulmonary function at
baseline, early onset, atopic” (31%)- “Moderate asthmatics and the majority had good lung
- “Confident and self-managing” (26%)- “Confident and accepting of their asthma” (35%)- “Confident but dependent on others” (6%)- “Concerned but confident in their health care professional”
(28%)- “Not confident in themselves on their health care
professional” (6%)
Diagnostics 2021, 11, 644 59 of 63
Table A2. Cont.
Study ID (Author, Year) Label
Factor Analysis
Alves, 2008 [18]
- “Treatment-resistant, more nocturnal symptoms andexacerbations” (32%)
- “Persistent airflow limitation, lower FEV1/FVC ratios in theinitial evaluation, more advanced age and longer durationof the disease” (55%)
- “Allergic rhinosinusitis, nonsmokers and reversible airflowobstruction” (48%)
- “Aspirin intolerance associated with near-fatal asthmaepisodes” (17%)
Moore, 2014 [52]
- “Mild-to-moderate early onset allergic asthma withpaucigranulocytic or eosinophilic sputum inflammatory cellpatterns” (31%)
- “Mild-to-moderate early onset allergic asthma withpaucigranulocytic or eosinophilic sputum inflammatory cellpatterns, OCS use” (30%)
- “Moderate-to-severe asthma with frequent health care usedespite treatment with high doses of inhaled or oralcorticosteroids, normal lung function” (28%)
- “Moderate-to-severe asthma with frequent health care usedespite treatment with high doses of inhaled or oralcorticosteroids, reduced lung function” (11%)
A.; et al. Systems medicine and integrated care to combat chronic noncommunicable diseases. Genome Med. 2011, 3, 43. [CrossRef][PubMed]
5. Padem, N.; Saltoun, C. Classification of asthma. Allergy Asthma Proc. 2019, 40, 385–388. [CrossRef]6. Basile, A.O.; Ritchie, M.D. Informatics and machine learning to define the phenotype. Expert Rev. Mol. Diagn. 2018, 18, 219–226.
[CrossRef]7. Lanza, S.T.; Collins, L.M.; Lemmon, D.R.; Schafer, J.L. PROC LCA: A SAS Procedure for Latent Class Analysis. Struct. Equ. Model.
A Multidiscip. J. 2007, 14, 671–694. [CrossRef]8. Loureiro, C.; Sa-Couto, P.; Todo-Bom, A.; Bousquet, J. Cluster analysis in phenotyping a Portuguese population. Rev. Port. de
et al. Validated and longitudinally stable asthma phenotypes based on cluster analysis of the ADEPT study. Respir. Res. 2016,17, 1–21. [CrossRef]
10. Wu, W.; Bleecker, E.; Moore, W.; Busse, W.W.; Castro, M.; Chung, K.F.; Calhoun, W.J.; Erzurum, S.; Gaston, B.; Israel, E.; et al.Unsupervised phenotyping of Severe Asthma Research Program participants using expanded lung data. J. Allergy Clin. Immunol.2014, 133, 1280–1288. [CrossRef]
11. Liberati, A.; Altman, D.G.; Tetzlaff, J.; Mulrow, C.; Gøtzsche, P.C.; Ioannidis, J.P.; Clarke, M.; Devereaux, P.J.; Kleijnen, J.; Moher, D.The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions:Explanation and elaboration. BMJ 2009, 339, b2700. [CrossRef]
12. Santos, C.M.D.C.; Pimenta, C.A.D.M.; Nobre, M.R.C. The PICO strategy for the research question construction and evidencesearch. Rev. Lat. Am. Enferm. 2007, 15, 508–511. [CrossRef] [PubMed]
13. Sterne, J.A.; Hernán, M.A.; Reeves, B.C.; Savovic, J.; Berkman, N.D.; Viswanathan, M.; Henry, D.; Altman, D.G.; Ansari, M.T.;Boutron, I.; et al. ROBINS-I: A tool for assessing risk of bias in non-randomised studies of interventions. BMJ 2016, 355, i4919.[CrossRef] [PubMed]
17. Agache, I.; Ciobanu, C. Risk Factors and Asthma Phenotypes in Children and Adults with Seasonal Allergic Rhinitis. PhysicianSportsmed. 2010, 38, 81–86. [CrossRef]
18. Alves, R.D.S.A.; Vianna, F.D.A.F.; Pereira, C.A.D.C. Fenótipos clínicos de asma grave. J. Bras. de Pneumol. 2008, 34, 646–653.[CrossRef]
19. Amaral, R.; Pereira, A.M.; Jacinto, T.; Malinovschi, A.; Janson, C.; Alving, K.; Fonseca, J.A. Comparison of hypothesisanddata-driven asthma phenotypes in NHANES 2007–2012: The importance of comprehensive data availability. Clin. Transl. Allergy2019, 9, 1–7. [CrossRef]
20. Amaral, R.; Bousquet, J.; Pereira, A.M.; Araújo, L.M.; Sá-Sousa, A.; Jacinto, T.; Almeida, R.; Delgado, L.; Fonseca, J.A. Disentanglingthe heterogeneity of allergic respiratory diseases by latent class analysis reveals novel phenotypes. Allergy 2019, 74, 698–708.[CrossRef]
21. Amelink, M.; De Nijs, S.B.; De Groot, J.C.; Van Tilburg, P.M.B.; Van Spiegel, P.I.; Krouwels, F.H.; Lutter, R.; Zwinderman, A.H.;Weersink, E.J.M.; Brinke, A.T.; et al. Three phenotypes of adult-onset asthma. Allergy 2013, 68, 674–680. [CrossRef] [PubMed]
23. Belhassen, M.; Langlois, C.; Laforest, L.; Dima, A.L.; Ginoux, M.; Sadatsafavi, M.; Van Ganse, E. Level of Asthma ControllerTherapy Before Admission to the Hospital. J. Allergy Clin. Immunol. Pract. 2016, 4, 877–883. [CrossRef] [PubMed]
24. Bochenek, G.; Kuschill-Dziurda, J.; Szafraniec, K.; Plutecka, H.; Szczeklik, A.; Nizankowska-Mogilnicka, E. Certain subphenotypesof aspirin-exacerbated respiratory disease distinguished by latent class analysis. J. Allergy Clin. Immunol. 2014, 133, 98–103.[CrossRef]
25. Boudier, A.; Curjuric, I.; Basagaña, X.; Hazgui, H.; Anto, J.M.; Bousquet, J.; Bridevaux, P.-O.; Dupuis-Lozeron, E.; Garcia-Aymerich,J.; Heinrich, J.; et al. Ten-Year Follow-up of Cluster-based Asthma Phenotypes in Adults. A Pooled Analysis of Three Cohorts. Am. J.Respir. Crit. Care Med. 2013, 188, 550–560. [CrossRef]
26. Chanoine, S.; Pin, I.; Sanchez, M.; Temam, S.; Pison, C.; Le Moual, N.; Severi, G.; Boutron-Ruault, M.-C.; Fournier, A.; Bousquet, J.;et al. Asthma Medication Ratio Phenotypes in Elderly Women. J. Allergy Clin. Immunol. Pract. 2018, 6, 897–906. [CrossRef]
34. Hsiao, H.-P.; Lin, M.-C.; Wu, C.-C.; Wang, C.-C.; Wang, T.-N. Sex-Specific Asthma Phenotypes, Inflammatory Patterns, andAsthma Control in a Cluster Analysis. J. Allergy Clin. Immunol. Pract. 2019, 7, 556–567.e15. [CrossRef]
35. Ilmarinen, P.; Tuomisto, L.E.; Niemelä, O.; Tommola, M.; Haanpää, J.; Kankaanranta, H. Cluster Analysis on Longitudinal Data ofPatients with Adult-Onset Asthma. J. Allergy Clin. Immunol. Pract. 2017, 5, 967–978. [CrossRef]
36. Jang, A.S.; Kwon, H.-S.; Cho, Y.S.; Bae, Y.J.; Kim, T.B.; Park, J.S.; Park, S.W.; Uh, S.-T.; Choi, J.-S.; Kim, Y.-H.; et al. Identification ofSubtypes of Refractory Asthma in Korean Patients by Cluster Analysis. Lung 2012, 191, 87–93. [CrossRef]
37. Janssens, T.; Verleden, G.; Bergh, O.V.D. Symptoms, Lung Function, and Perception of Asthma Control: An Exploration into theHeterogeneity of the Asthma Control Construct. J. Asthma 2011, 49, 63–69. [CrossRef]
38. Jeong, A.; Imboden, M.; Hansen, S.; Zemp, E.; Bridevaux, P.-O.; Lovison, G.; Schindler, C.; Probst-Hensch, N. Heterogeneity ofobesity-asthma association disentangled by latent class analysis, the SAPALDIA cohort. Respir. Med. 2017, 125, 25–32. [CrossRef]
39. Khusial, R.J.; for the ACCURATE Study Group; Sont, J.K.; Loijmans, R.J.B.; Snoeck-Stroband, J.B.; Assendelft, P.J.J.; Schermer,T.R.J.; Honkoop, P.J.; ACCURATE Study Group. Longitudinal outcomes of different asthma phenotypes in primary care, anobservational study. NPJ Prim. Care Respir. Med. 2017, 27, 55. [CrossRef]
40. Kim, J.-H.; Chang, H.S.; Shin, S.W.; Baek, D.G.; Son, J.-H.; Park, C.-S.; Park, J.-S. Lung Function Trajectory Types in Never-SmokingAdults with Asthma: Clinical Features and Inflammatory Patterns. Allergy Asthma Immunol. Res. 2018, 10, 614–627. [CrossRef]
41. Kim, M.-A.; Shin, S.-W.; Park, J.-S.; Uh, S.-T.; Chang, H.S.; Bae, D.-J.; Cho, Y.S.; Park, H.-S.; Yoon, H.J.; Choi, B.W.; et al. ClinicalCharacteristics of Exacerbation-Prone Adult Asthmatics Identified by Cluster Analysis. Allergy Asthma Immunol. Res. 2017,9, 483–490. [CrossRef]
42. Kim, T.-B.; Jang, A.-S.; Kwon, H.-S.; Park, J.-S.; Chang, Y.-S.; Cho, S.-H.; Choi, B.W.; Park, J.-W.; Nam, D.-H.; Yoon, H.-J.; et al.Identification of asthma clusters in two independent Korean adult asthma cohorts. Eur. Respir. J. 2012, 41, 1308–1314. [CrossRef][PubMed]
43. Kisiel, M.A.; Zhou, X.; Sundh, J.; Ställberg, B.; Lisspers, K.; Malinovschi, A.; Sandelowsky, H.; Montgomery, S.; Nager, A.; Janson,C. Data-driven questionnaire-based cluster analysis of asthma in Swedish adults. NPJ Prim. Care Respir. Med. 2020, 30, 14–19.[CrossRef] [PubMed]
44. Konno, S.; Taniguchi, N.; Makita, H.; Nakamaru, Y.; Shimizu, K.; Shijubo, N.; Fuke, S.; Takeyabu, K.; Oguri, M.; Kimura, H.; et al.Distinct Phenotypes of Cigarette Smokers Identified by Cluster Analysis of Patients with Severe Asthma. Ann. Am. Thorac. Soc.2015, 12, 1771–1780. [CrossRef] [PubMed]
45. Konstantellou, E.; Papaioannou, A.I.; Loukides, S.; Patentalakis, G.; Papaporfyriou, A.; Hillas, G.; Papiris, S.; Koulouris, N.;Bakakos, P.; Kostikas, K. Persistent airflow obstruction in patients with asthma: Characteristics of a distinct clinical phenotype.Respir. Med. 2015, 109, 1404–1409. [CrossRef]
46. Labor, M.; Labor, S.; Juric, I.; Fijacko, V.; Grle, S.P.; Plavec, D. Mood disorders in adult asthma phenotypes. J. Asthma 2017,55, 57–65. [CrossRef]
47. Lee, J.H.; Rhee, C.K.; Kim, K.; Kim, J.-A.; Kim, S.H.; Yoo, K.-H.; Kim, W.J.; Park, Y.B.; Park, H.Y.; Jung, K.-S. Identification ofsubtypes in subjects with mild-to-moderate airflow limitation and its clinical and socioeconomic implications. Int. J. ChronicObstr. Pulm. Dis. 2017, 12, 1135–1144. [CrossRef]
48. Lefaudeux, D.; De Meulder, B.; Loza, M.J.; Peffer, N.; Rowe, A.; Baribaud, F.; Bansal, A.T.; Lutter, R.; Sousa, A.R.; Corfield, J.; et al.U-BIOPRED clinical adult asthma clusters linked to a subset of sputum omics. J. Allergy Clin. Immunol. 2017, 139, 1797–1807.[CrossRef]
49. Lemiere, C.; Nguyen, S.; Sava, F.; D’Alpaos, V.; Huaux, F.; Vandenplas, O. Occupational asthma phenotypes identified byincreased fractional exhaled nitric oxide after exposure to causal agents. J. Allergy Clin. Immunol. 2014, 134, 1063–1067. [CrossRef]
50. Mäkikyrö, E.M.S.; Jaakkola, M.S.; Jaakkola, J.J.K. Subtypes of asthma based on asthma control and severity: A latent class analysis.Respir. Res. 2017, 18, 1–11. [CrossRef]
51. Moore, W.C.; Meyers, D.A.; Wenzel, S.E.; Teague, W.G.; Li, H.; Li, X.; D’Agostino, R., Jr.; Castro, M.; Curran-Everett, D.; Fitzpatrick,A.M.; et al. Identification of Asthma Phenotypes Using Cluster Analysis in the Severe Asthma Research Program. Am. J. Respir.Crit. Care Med. 2010, 181, 315–323. [CrossRef]
52. Moore, W.C.; Hastie, A.T.; Li, X.; Li, H.; Busse, W.W.; Jarjour, N.N.; Wenzel, S.E.; Peters, S.P.; Meyers, D.A.; Bleecker, E.R. Sputumneutrophil counts are associated with more severe asthma phenotypes using cluster analysis. J. Allergy Clin. Immunol. 2014,133, 1557–1563.e5. [CrossRef]
53. Musk, A.W.; Knuiman, M.; Hunter, M.; Hui, J.; Palmer, L.J.; Beilby, J.; Divitini, M.; Mulrennan, S.; James, A. Patterns of airwaydisease and the clinical diagnosis of asthma in the Busselton population. Eur. Respir. J. 2011, 38, 1053–1059. [CrossRef]
54. Nagasaki, T.; Matsumoto, H.; Kanemitsu, Y.; Izuhara, K.; Tohda, Y.; Kita, H.; Horiguchi, T.; Kuwabara, K.; Tomii, K.; Otsuka, K.;et al. Integrating longitudinal information on pulmonary function and inflammation using asthma phenotypes. J. Allergy Clin.Immunol. 2014, 133, 1474–1477.e2. [CrossRef]
55. Newby, C.; Heaney, L.G.; Menzies-Gow, A.; Niven, R.M.; Mansur, A.; Bucknall, C.; Chaudhuri, R.; Thompson, J.; Burton, P.;Brightling, C.; et al. Statistical Cluster Analysis of the British Thoracic Society Severe Refractory Asthma Registry: ClinicalOutcomes and Phenotype Stability. PLoS ONE 2014, 9, e102987. [CrossRef]
56. Oh, J.H.; Ahn, K.-M.; Chung, S.J.; Shim, J.-S.; Park, H.-W. Usefulness of routine blood test-driven clusters for predicting acuteexacerbation in patients with asthma. Respir. Med. 2020, 170, 170. [CrossRef]
57. Park, H.-W.; Song, W.-J.; Kim, S.-H.; Park, H.-K.; Kim, S.-H.; Kwon, Y.E.; Kwon, H.-S.; Kim, T.-B.; Chang, Y.-S.; Cho, Y.-S.; et al.Classification and implementation of asthma phenotypes in elderly patients. Ann. Allergy Asthma Immunol. 2015, 114, 18–22.[CrossRef]
58. Park, S.Y.; Baek, S.; Kim, S.; Yoon, S.-Y.; Kwon, H.-S.; Chang, Y.-S.; Cho, Y.S.; Jang, A.-S.; Park, J.W.; Nahm, N.-H.; et al. ClinicalSignificance of Asthma Clusters by Longitudinal Analysis in Korean Asthma Cohort. PLoS ONE 2013, 8, e83540. [CrossRef][PubMed]
59. Park, S.Y.; Jung, H.W.; Lee, J.M.; Shin, B.; Kim, H.J.; Kim, M.-H.; Song, W.-J.; Kwon, H.-S.; Jung, J.-W.; Kim, S.-H.; et al. NovelTrajectories for Identifying Asthma Phenotypes: A Longitudinal Study in Korean Asthma Cohort, COREA. J. Allergy Clin.Immunol. Pr. 2019, 7, 1850–1857.e4. [CrossRef] [PubMed]
60. Qiu, R.; Xie, J.; Chung, K.F.; Li, N.; Yang, Z.; He, M.; Li, J.; Chen, R.; Zhong, N.; Zhang, Q. Asthma Phenotypes Defined fromParameters Obtained During Recovery From a Hospital-Treated Exacerbation. J. Allergy Clin. Immunol. Pract. 2018, 6, 1960–1967.[CrossRef] [PubMed]
61. Rakowski, E.; Zhao, S.; Liu, M.; Ahuja, S.; Durmus, N.; Grunig, G.; De Lafaille, M.C.; Wu, Y.; Reibman, J. Variability of bloodeosinophils in patients in a clinic for severe asthma. Clin. Exp. Allergy 2018, 49, 163–170. [CrossRef] [PubMed]
62. Rootmensen, G.; Van Keimpema, A.; Zwinderman, A.; Sterk, P. Clinical phenotypes of obstructive airway diseases in an outpatientpopulation. J. Asthma 2016, 53, 1026–1032. [CrossRef] [PubMed]
63. Sakagami, T.; Hasegawa, T.; Koya, T.; Furukawa, T.; Kawakami, H.; Kimura, Y.; Hoshino, Y.; Sakamoto, H.; Shima, K.; Kagamu,H.; et al. Cluster analysis identifies characteristic phenotypes of asthma with accelerated lung function decline. J. Asthma 2013, 51,113–118. [CrossRef] [PubMed]
64. Schatz, M.; Hsu, J.-W.Y.; Zeiger, R.S.; Chen, W.; Dorenbaum, A.; Chipps, B.E.; Haselkorn, T. Phenotypes determined by clusteranalysis in severe or difficult-to-treat asthma. J. Allergy Clin. Immunol. 2014, 133, 1549–1556. [CrossRef] [PubMed]
71. Tanaka, H.; Nakatani, E.; Fukutomi, Y.; Sekiya, K.; Kaneda, H.; Iikura, M.; Yoshida, M.; Takahashi, K.; Tomii, K.; Nishikawa, M.;et al. Identification of patterns of factors preceding severe or life-threatening asthma exacerbations in a nationwide study. Allergy2018, 73, 1110–1118. [CrossRef] [PubMed]
72. Tay, T.R.; Choo, X.N.; Yii, A.; Chung, K.F.; Chan, Y.H.; Wong, H.S.; Chan, A.; Tee, A.; Koh, M.S. Asthma phenotypes in amulti-ethnic Asian cohort. Respir. Med. 2019, 157, 42–48. [CrossRef] [PubMed]
73. Van Der Molen, T.; Fletcher, M.; Price, D. Identifying Patient Attitudinal Clusters Associated with Asthma Control: The EuropeanREALISE Survey. J. Allergy Clin. Immunol. Pract. 2018, 6, 962–971. [CrossRef] [PubMed]
74. Wang, L.; Liang, R.; Zhou, T.; Zheng, J.; Liang, B.M.; Zhang, H.P.; Luo, F.M.; Gibson, P.G.; Wang, G. Identification and validationof asthma phenotypes in Chinese population using cluster analysis. Ann. Allergy Asthma Immunol. 2017, 119, 324–332. [CrossRef]
75. Weatherall, M.; Travers, J.; Shirtcliffe, P.M.; Marsh, S.E.; Williams, M.V.; Nowitz, M.R.; Aldington, S.; Beasley, R. Distinct clinicalphenotypes of airways disease defined by cluster analysis. Eur. Respir. J. 2009, 34, 812–818. [CrossRef]
76. Wu, D.; Bleier, B.S.; Li, L.; Zhan, X.; Zhang, L.; Lv, Q.; Wang, J.; Wei, Y. Clinical Phenotypes of Nasal Polyps and ComorbidAsthma Based on Cluster Analysis of Disease History. J. Allergy Clin. Immunol. Pract. 2018, 6, 1297–1305.e1. [CrossRef]
77. Ye, W.-J.; Xu, W.-G.; Guo, X.-J.; Han, F.-F.; Peng, J.; Li, X.-M.; Guan, W.-B.; Yu, L.-W.; Sun, J.-Y.; Cui, Z.-L.; et al. Differencesin airway remodeling and airway inflammation among moderate-severe asthma clinical phenotypes. J. Thorac. Dis. 2017, 9,2904–2914. [CrossRef] [PubMed]
78. Youroukova, V.M.; Dimitrova, D.G.; Valerieva, A.D.; Lesichkova, S.S.; Velikova, T.V.; Ivanova-Todorova, E.I.; Tumangelova-Yuzeir,K.D. Phenotypes Determined by Cluster Analysis in Moderate to Severe Bronchial Asthma. Folia Med. 2017, 59, 165–173.[CrossRef] [PubMed]
79. Zaihra, T.; Walsh, C.J.; Ahmed, S.; Fugère, C.; Hamid, Q.A.; Olivenstein, R.; Martin, J.G.; Benedetti, A. Phenotyping of difficultasthma using longitudinal physiological and biomarker measurements reveals significant differences in stability between clusters.BMC Pulm. Med. 2016, 16, 74. [CrossRef]
80. Cuschieri, S. The STROBE guidelines. Saudi J. Anaesth. 2019, 13, S31–S34. [CrossRef]81. Wenzel, S.E. Asthma phenotypes: The evolution from clinical to molecular approaches. Nat. Med. 2012, 18, 716–725. [CrossRef]